Along with development of e-commerce, online shopping becomes a consumption pattern for more and more people. Purchase of clothes is one of main consumption directions for the people. Since the clothes are generally required to be tried on, a virtual fitting technique emerges as the times require. Virtual fitting refers to using computer technology to let a virtual model try on the clothes sold on line instead of an actual user, and a certain reference is given to the user to selectively purchase the clothes on line by means of an effect presented by the fitting of the virtual model, which facilitates the purchase of proper clothes by the user. In the virtual fitting, a piece of clothing used is required to be processed, and in addition to performing image processing to a clothing image, it is further required to acquire some attribute points on the clothing image, the attribute points correspond to points on the virtual fitting model, so that a virtual fitting software can accurately make the virtual fitting model “put on” the piece of clothing.
At present, a manner of acquiring clothing image attribute points mainly depends on manpower, i.e., a person clicks a position to be calibrated on the clothing image to form an attribute point at this position. The clothing image is generally a front view of the piece of clothing, and at this time, the clothing image is bilaterally symmetric. There are generally more than ten attribute points of the clothing image, and a manner of manual calibration is comparatively time-consuming and comparatively low in efficiency.